The Collaboratory for the Study of Earthquake Predictability version 2.0 (CSEP2.0): New Capabilities in Earthquake Forecasting and Testing

William H. Savran, Philip J. Maechling, Maximilian J. Werner, Thomas H. Jordan, Danijel Schorlemmer, David A. Rhoades, Warner Marzocchi, John Yu, & John E. Vidale

Published August 15, 2018, SCEC Contribution #8805, 2018 SCEC Annual Meeting Poster #033

The Collaboratory for the Study of Earthquake Predictability (CSEP) supports an international effort to rigorously conduct earthquake forecasting experiments. CSEP testing centers now operate in California, New Zealand, Japan, China, and Europe, with 442 models under prospective evaluation. CSEP supports retrospective forecast evaluations (testing forecasts on existing data) and prospective experiments (testing forecasts on data to be collected in the future). Current CSEP experiments evaluate forecasts that are expressed as expected rates in small space-magnitude bins and updated at regular intervals (e.g., daily or yearly). This experiment design is simple and allows a wide range of models to participate. However, recent forecast models, including candidate models for Operational Earthquake Forecasting (OEF), involve simulation of thousands of synthetic seismicity catalogs (stochastic event sets), which express important dependency structures between triggered earthquakes. In addition, some models, such as the third version of the Uniform California Earthquake Rupture Forecast (UCERF3), forecast spatially extensive ruptures on specified faults. As part of CSEP’s second phase (CSEP2.0), we are redesigning CSEP’s software to support the testing of such models. Requirements include access to high-performance computing, distributed processing of forecasts and evaluations, and streamlined data management, as well as adhering to CSEP’s goals of transparency and reproducibility within a controlled, open-source software environment. This redesign is being implemented through a modularization of CSEP’s codes to allow for tailored experiments that do not fit into the current framework allowing for easy recombination of modules for new types of experiments should greatly expand CSEP’s model and experiment space. Experiments will be described using directed-acyclic graphs (DAGs) that allow us to execute the computation using workflow tools. This allows for introducing high-performance computing resources while still maintaining the core values of CSEP testing centers. The improved capabilities of the CSEP software will be made available to the earthquake forecast research community. We follow best-practices of open-source software, which include continuous integration of the scientific and utilities codes, a collection of unit and acceptance tests, and thorough documentation that includes example forecasts and evaluations.

Key Words
earthquake forecasting, scientific workflows, csep, reproducibility

Citation
Savran, W. H., Maechling, P. J., Werner, M. J., Jordan, T. H., Schorlemmer, D., Rhoades, D. A., Marzocchi, W., Yu, J., & Vidale, J. E. (2018, 08). The Collaboratory for the Study of Earthquake Predictability version 2.0 (CSEP2.0): New Capabilities in Earthquake Forecasting and Testing . Poster Presentation at 2018 SCEC Annual Meeting.

Related Projects & Working Groups
Collaboratory for the Study of Earthquake Predictability (CSEP)